Browsing by Author "Hsieh, Jenwei"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Efficient Implementation of RAID-5 Using Disk Based Read Modify Writes
    (1997) Shim, Sangyup; Wang, Yuewei; Hsieh, Jenwei; Chang, Taisheng; Du, David H.C.
    Redundant Array of Inexpensive Disks (RAID) is often used to provide a fault tolerance capability for disk failures in database systems. An efficient implementation of small writes is the most important issue to achieve high throughput because most of writes in databases are small. The tradittional RAID implementation requires a RAID controller to construct parity blocks for writes (host based writes). An RAID implementation based on an exclusive-or (xor) engine in each disk may minimize link traffic and improve the response times for small writes. This is called disk based writes. However, many implementation issues have to resolved before the higher performance can be realized. We have observed that a straightforward implementation of disk based writes achieves only half of the potential throughput because of the interactions among commands. One of the implementation issues is to prevent a state of deadlock. A state of deadlock may occur when multiple disk based writes are executed simultaneously. The implementation issues addressed in this paper include how to prevent a state of deadlock, how to enhance cache replacement policy in disks, how to increase concurrency among disk basedd writes, and how to minimize the interactions of many commands. This paper investigates the effectiveness of disk based writes, and shows the performance comparison between a disk based write and a host based write in RAIDs. In a stress test, the aggregate throughput was increased, and the average command latency was reduced.
  • Thumbnail Image
    Item
    High-Speed Network Support for High-Performance Network Computing and Multimedia Communications
    (1997) Hsieh, Jenwei
    The prevalence of computer networks has shifted the computing paradigm from mainframe or host-centric computing to network-centric computing. In networkcentric computing, applications are executed distributedly on a collection of computers interconnected via local and wide area networks. The performance of networkcentric applications can be dramatically improved with switch-based high-speed networks, such as HIPPI, ATM, and Fibre Channel. In this study, we focus on the high-speed network support for two important applications in network-centric computing: high-performance network computing and multimedia communication. One important class of network computing is cluster computing, which enables a collection of locally interconnected computers to be used as a general-purpose parallel computing system. Large problems can be solved cost effectively by using the aggregate processing power and memory space of a cluster. However, communication between processors has long been the bottleneck of cluster computing. We have especially concentrated on maximizing the achievable throughput and minimizing the communication delay for cluster computing in homogeneous environments. We have enhanced a popular cluster computing environment, Parallel Virtual Machine (PVM) with clusters of workstations on either local ATM or HIPPI networks. One possible extension of cluster computing is to incorporate clusters of computers via wide area networks. This is called meta-computing. For example, a group of diverse high-performance computers from several geographically distributed supercomputer centers can be employed to solve large problems. ATM is the de facto standard for wide area networks. However, most of the supercomputer centers use HIPPI in their computing facilities. The internetworking of HIP PI networks and wide area ATM networks becomes an important issue for met a-computing. Two feasible solutions for the problem, HIPP! Tunneling and IP Routing, have been studied in this thesis. Multimedia communication imposes another challenge for high speed networks. The delivery of continuous media requires high communication bandwidth and realtime constraint. Ve have studied two new CBR transmission schemes, called PCRassist CBR (PCBR) and PCR-assist Dual-Rate CBR (PDCBR), which employ the Program Clock References (PCR) embedded in the MPEG-2 Transport Streams to regulate their transmission. The two schemes provide flexible trade-off between buffer requirement and transmission rates.